Turbidity in Sugar Products - American Chemical Society

golden brown color appears. Estimations are conveniently carried out in batches of eight. The ceric sulfate used is standardized against a pure grade ...
0 downloads 0 Views 642KB Size
MAY 15, 1937

ANALYTICAL EDITION

Eight drops of Setopaline C indicator solution are then added and the titration is continued to a sharp change from golden brown to a light yellow color. A 0.01 N solution of ceric sulfate is made up for the use of sugar analysis by diluting 20 cc. of the 0.25 N stock solution and 50 cc. of 1 to 1 sulfuric acid to 500 cc. The 0.01 N ceric sulfate solution may also be standardized against Mohr's salt. A 0.01 N standard solution of the ferrous ammonium sulfate should be used for this purpose. The stock solution is kept in a cool place and the dilute 0.01 N ceric sulfate is made up as needed.

Procedure The plant extract is clarified as previously described (1). Five cubic centimeters of the Clarified solution containing not more than 3.5 mg. of reducing sugar are mixed with 5 cc. of the alkaline ferricyanide in a 145 X 28 mm. Pyrex glass test tube. The tube is heated in a boiling water bath or immersed in a steam bath and heated for exactly 15 minutes. The tube with the contents is then cooled to room temperature by immersing in running water for about 3 minutes. Five cubic centimeters of 5 N sulfuric acid are added and the contents are mixed by shaking the tube. Seven to 10 drops of the Setopaline C indicator are introduced and titrated with the 0.01 N ceric sulfate from a 10-cc. buret until a golden brown color a pears. Estimations are conveniently carried out in batches ofeight.

229

The ceric sulfate used is standardized against a pure grade of glucose by the same procedure, using 5-cc. aliquots of solution containing known amounts of the sugar. From the titration of pure glucose the number of cubic centimeters of ceric sulfate equivalent to 1 mg. of glucose is calculated. It was found that 3.00 cc. of 0.01 N ceric sulfate solution are equivalent to 1 mg. of glucose. It is convenient to have on hand a solution of pure 1 per cent glucose made up in 0.25 per cent benzoic acid. When prepared in this way the glucose will keep for several months. Whenever needed, a few cubic centimeters of this standard solution are diluted to the proper strength and used for standardization of the dilute ceric sulfate solution. Standardization of the ceric sulfate solution against glucose is necessary whenever it is made from the 0.25 N stock solution.

Literature Cited (1) Hassid, W. Z., IND.ENO.CEEM.,Anal. Ed., 8,138 (1936).

RBCEIVED March 15, 1937.

Turbidity in Sugar Products V. Color and Turbidity of Hard Refined Sugars F. W. ZERBAN

AND

LOUIS SATTLER, New York Sugar Trade Laboratory, New York, N. Y.

T

H E previous two articles of this series (3, 6) dealt with the determination of coloring matter and turbidity in raw cane sugars. This investigation has now been extended to hard refined sugars. Of the twenty-nine samples examined, twenty-six were granulated cane sugars, two were powdered cane sugars, and one was a beet sugar. Three analytical procedures were used in this work: (1) Balch; (2) Zerban, Sattler and Lorge; (3) Landt and Witte.

Balch Method In the method of Balch (1) a 60-Brix solution of the sugar is prepared, and a portion of it is filtered. The unfiltered solution is read in the spectrophotometer against the filtered solution a t one or more wave lengths, and the result is expressed in terms of the specific absorptive index, -log t . The concentration of the coloring matter is found by reading the filtered solution against water, and converting the result also into -log t. In applying this method to white sugars, the writers used the same filtering materials as before: Celite analytical filter aid, specially prepared asbestos, and purified silica gel. Since hard refined sugars contain very little coloring matter and turbidity, long absorption cells must be employed. The cells used by the writers were 100 mm. in length, the maximum available for the Keuffel and Esser spectrophotometer. Still longer cells would be preferable. The readings were taken a t wave length 529 mp, in order to permit direct comparison with the measurements made according to the other two procedures.

Zerban, Sattler, and Lorge Method In the method of Zerban, Sattler, and Lorge (4) the transmittancy, T, and the Tyndall beam intensity, R, of the unfiltered 60-Brix solution are measured with the Pulfrich photometer in a 2.5-mm. cell. R is expressed in per cent of the Tyndall beam intensity of the standard turbid glass block

of the instrument. Color filters are used for measurements a t various wave lengths. The concentration of turbidity, N , and of coloring matter C, are then found by means of Equations 1 and 2. R = aNlc-c

- log T = N

+C

The numerical values of the constants a and k are given in the paper referred to (4). In practice it is easier to find N and C from a graph based on the two equations rather than by calculation. N and C are thus expressed in terms of -log T for a 60-Brix solution and a cell thickness of 2.5 mm. I n refined white sugars the color is generally so close to zero that the expression k-C in Equation 1 is almost equal to unity, and N becomes directly proportional to R, or nearly so. An attempt was made to measure the Tyndall beam intensity of the refined sugar solutions in the cells of 2.45- and 2.46-mm. thickness, previously used in raw sugar work, but the turbidity was so small and the correction for the turbidity of the water used for dissolving the sugar was relatively so large that reliable results could not be obtained. The measurements were therefore repeated in a 26-mm. Zechmeister beaker, the effective depth of which is 16.3 mm. The transmittancy of the unfiltered solutions was measured in a 100-mm. tube with the spectrophotometer, the same figure being used as in the method of Balch. The results for -log TI00 and for R18.3are shown in Table I, columns 2 and 3, respectively, for wave length 529 mp, I n order to calculate N and C from -log T and R by means of Equations 1 and 2, it was necessary first to reduce -log 2'10 and R16.a to a depth of 2.455 mm. It was found that over such a wide range of thickness the optical density, D (= -log T), is not directly proportional to the depth, b, but a power function of it, according to the equation

DI: A

5

(bi h)"

(3)

230

INDUSTRIAL AND ENGINEERING CHEMISTRY

VOL. 9, NO. 5

The -log T values for the total absorption and for that due to turbidity alone, obtained by Balch’s method with 60Brix solutions and a 100-mm. tube, were reduced to 1-em. thickness by multiplying by 0.11092 (0.10.965), and to a concentration of TABLEI. EXPERIMENTAL DATAAND CALCULAZED VALUES FOR METHODS OF ZERBAN, 100 grams in 100 ml. by multiplying SATTLER, AND LORQE, AND OF LANDT AND WITTE by 1.296. Beer’s law was assumed 1 2 3 4 8 6 7 8 9 to hold within the limits of error, Absolute N, C fk for the increase in concentration was -log T, R 100 A , -log T. log R only about 30 per cent. The conNo. 100 mm. 16.3 mm. 2.455 mm. 2.455 m h . 2.455 mm. 2.458)mm. 16.3 k m . Witte 0 0023 centration of coloring matter was 1.0811 0.0030 1.106 0.0021 0.1767 0.0051 73.49 1.171 0.0031 1.4694 0.0049 179.70 0.2746 0.0080 then found by deducting the turo.oo20 0.0059 1.098 1.0334 0,0029 0.0018 65.85 0.1634 0.0047 1.026 0.6228 0,0007 0.0005 0 0007 bidity, expressed as -log t, from the 0.0012 25.58 0.0416 1.053 0.5850 0.0006 0.0020 0.0026 23.45 0.0901 0 0045 0.0007 total -log t of the unfiltered soh1.149 1.3563 0.0038 0.0032 0.0070 0.2409 138.49 1.086 0 0042 oo12 tion. The turbidity and c o l o r i n g 1.3465 0.0003 0.0038 0.0041 0.1428 135.40 1.059 0.0017 0.8173 0.0012 0.0029 40.03 0.0992 8 1.079 0,0023 0 0017 matter, calculated in this manner, 0.9587 0.0015 0.0038 65.44 0.1307 9 0 0031 1.1913 1.160 0.0048 0.0027 0.0075 94.71 0.2577 10 are shown in Table 11, columns 2 1.049 0.0014 oolo 0.7621 0.0010 0.0832 36.26 0.0024 11 1.062 0.9961 0.0013 0 0018 to 7, for filtration with Celite, as0.0017 0.1036 60.43 0.0030 12 1.060 0.9460 0.0016 0.0015 0.1025 53.85 0.0030 13 0 0024 bestos, and silica gel, respectively. 1.277 1.0312 0.0104 0.0019 0.4250 0.0123 65.51 14 1.094 1.4246 0.0044 0.0002 162.10 0.1574 0.0046 16 0 0050 The -log T and N values, ob1,080 0 0029 1.1984 0.0028 0.0011 0.0039 96.28 0.1331 16 1.075 1.2264 0.0029 0.0008 0 0031 tained by method 2 for a cell thick0.1262 0.0037 102.70 17 1.057 0 0018 0.0011 0.9845 0.0017 68.84 0.0965 0.0028 18 ness of 2.455 mm. and 60-Brix con1.100 0 0032 1.2215 0,0028 0.0020 0.0048 101.55 0.1662 19 1.053 0.9259 0.0014 0.0012 0.0026 61.41 0.0904 20 0.0015 centration, were similarly converted 1.0841 0.0002 1.044 0 0022 0.0020 0.0022 0.0766 74.00 21 into -log t , multiplying by 3.8238 1.054 0,0019 0 0021 1.0704 0.0008 0.0918 71.70 0.0027 22 1,025 0.0000 0.0013 0.8863 0.0013 0.0443 46.93 23 0014 (4.07340.955),and the product by 1 . 146 0.0005 00 0076 0.0064 1,5863 0.0069 235.21 0.2377 24 0.0027 1.10s 0.0025 1.1605 0.0052 88.23 0.1785 0 002s 1.296. The C values were again 26 0 0021 0.0008 1.0545 1.055 0.0019 0.0027 69.13 0.0935 26 found by difference. The N and C 1.081 0 0029 0.0012 1.1885 0.0027 0,0039 94.11 27 0.1347 1 118 0.0025 1.2669 0.0032 0.0057 112.73 0.1952 28 figures are given in Table 11,columns 0 0036 oo36 1.099 0.0015 1.2760 0.0033 0.0048 116.12 0.1638 29 8 and 9. T h e a b s o l u t e turbidities, found by the Landt and Witte method, were multiplied by 1.296, The values of -log T2.455and of R2.46S, calculated in this manner, to base them also on a concentration of 100 grams in 100 are given in columns 4 and 5 of Table 1. The corresponding ml., and the values thus calculated are shown in Table 11, values of N and C were next found from a greatly enlarged column 10. graph, similar to that shown in Figure 1 of Part I1 of this A comparison of the results obtained with the three filtering series (4), and the results are given in columns 6 and 7 of materials confirms the previous findings with raw sugars. Table I. The color concentration, arrived a t by asbestos filtration, averages higher than that found by Celite filtration, and Landt and Witte Method silica gel gives the lowest average color figures, but in many I n the method of Landt and Witte (2) the turbidity alone individual cases filtration with asbestos results in lower color is determined with the Pulfrich photometer, in the same values than filtration with Celite. Silica gel gives lower manner as in the procedure of Zerban, Sattler, and Lorge. color figures than asbestos in all cases but one, and lower ones The Tyndall beam intensity is expressed as A , which is equal than Celite in all but four. to 0.01 R. The so-called “absolute” turbidity is then calIn the preceding paper, dealing with raw sugars, it was culated by the formula shown that the discrepancies in the results obtained with the three filtering agents may be explained by their selective Absolute turbidity = A x f k X D X t (4) effect. All remove both turbidity and coloring matter, but to a varying degree, depending on particle size. The same is wherefh is a correction factor which is a function of the transtrue for white sugars. mittancy. The values of this factor are shown in Table I, Table I1 also shows that the method of Zerban, Sattler, column 8, for each of the 29 samples. D is a factor varying and Lorge, employing only unfiltered solutions for the with the depth of layer used. For the 26-mm. Zechmeister measurements, usually gives lower turbidity values and beaker D equals 1. t is the absolute turbidity of the standard correspondingly higher color values than the filtration glass block-i. e., the Tyndall beam intensity-for a 1-em. methods. I n a few individual cases the reverse is true. depth of layer, expressed as a fraction of the incident light Both higher and lower color values may be explained by the intensity, under the assumption that the primary light is selective effect of the filtering agents. In the former case completely scattered, and that there is no multiple reflection. the filtering material removes relatively more coloring matter The t of the glass block used in this investigation, for wave from the solution; in the latter a relatively large portion of length 529 mp, is 0.00282. The absolute turbidities of the the turbidity goes into the filtrate, and, causing absorption, sugar solutions, calculated by means of Equation4, are shown is registered as coloring matter. in Table I, column 9. It remains to inquire whether the methods utilizing the transmittancy and Tyndall beam intensity of unfiltered Comparative Results solutions give a reliable measure of the color and turbidity, for white sugars at least. Information on this point may be In order to make a direct comparison between the results obtained by a comparison between the results of the method obtained by methods 1 and 2, it is necessary to reduce them of Zerban, Sattler, and Lorge and that of Landt and Witte. all to a common basis, and the specific absorptive index, The latter method has a ,sound theoretical foundation when -log t, was chosenfor this purpose, The results of themethod coloring matter is absent or very small in quantity, as is the of Landt and Witte cannot be converted directly into -log t, case with white sugars. Under these conditions the turbidity but the relationship between the two will be discussed. The exponent n was found to equal 0.955. Consequently, -log TIMhad to be multiplied by 0.02901 to reduce it to -log T2.455. Similarly, Rie.ahad to be multiplied by 0.16401.

~$$~~‘

I

MAY 15, 1937

AXALYTICAL EDITION

231

concentration is practically proportional to the Tyndall beam intensity, but a correction must be applied for absorption.

1 and 2. It stands to reason that this condition is the exception rather than the rule, and it follows that the method of the writers, using unfiltered solutions, does not give absolute figures for color and turbidity TABLE 11. COMPARISON OF RESULTS WITH FILTERED AND UNFILTERED SOLUTIONS~ any more than the filtration 3 4 1 2 5 6 7 8 9 10 methods, and that the results Landt and are in both cases merely comCelite Aabestos Silica Gel Witte No. Turbidity Color Turbidity Color Turbidity Color N Turbidity parative. However, the time 1 0,0183 0.0104 0,0149 0.0030 required for an analyeis by the 2 0.0359 0.0242 0.0153 0.0077 3 0.0165 0.0035 0,0090 0.0025 0.0145 00.0026 0010 method of Zerban, Sattler, and 4 0.0041 5 0.0094 0.0030 0.0100 0,0009 Lorge is much shorter because 6 0.0275 0.0190 0.0156 0.0058 the tedious filtration process is 7 0.0179 0.0190 0.0015 0.0054 8 0.0098 0.0059 0.0074 0.0114 0.0084 0.0022 0.0016 avoided, while the number of 9 0.0162 10 0.0336 0.0134 0.0237 '0.0040 opticalobservations is the same 11 0.0077 0,0043 0.0031 0.0089 0.0096 0,0024 0.0050 0.0070 0.0014 in both methods. The only 12 0.0134 0,0016 0.0011 0.0138 0.0134 0.0016 0.0084 0.0065 0.0024 drawback of the method with13 0.0115 0,0032 0.0024 0,0123 0.0130 0.0017 0.0073 0.0073 0.0021 0.0094 14 0.0649 0.0062 0.0517 0.0555 0.0056 0.0094 0.0517 0.0031 out filtration is that the meas0.0058 0.0168 15 0.0196 0.0030 0.0210 0.0016 0.0216 0.0010 0.0065 0.0051 0.0140 16 0.0153 0.0038 0.0171 0.0020 0.0137 0.0054 0.0038 u r e m e n t s of Tyndall beam 0.0016 0.0027 17 0.0154 0.0165 0.0175 0.0006 0.0142 0.0039. 0.0040 0.0036 0.0026 18 0.0113 0.0103 0.0117 0,0022 0.0085 0.0054 0,0023 intensity must be made with 0.0064 0.0042 0.0175 19 0.0197 0.0203 0.0036 0.0137 0.0102 0.0041 the greatest care and attention 0.0060 0,0035 0.0070 20 0.0095 0.0114 0.0016 0.0070 0.0080 0,0020 0.0007 0.0030 21 0.0103 0.0080 0.0107 0,0003 0.0100 0.0010 0.0028 to details, as has been pointed 0.0005 0.0015 22 0.0127 0.0117 0.0130 0,0002 0.0093 0.0039 0.0028 out by Landt and Witte as well 0.0002 0.0029 0,0009 23 0.0062 0.0035 0.0055 0.0064 0.0000 0.0018 0.0034 0.0290 0 0051 0.0312 0.0029 24 0.0307 0.0316 0.0025 0.0099 as by the writers. 0.0037 25 0.0219 0.0214 0.0042 0.0243 0.0013 0.0123 0.0133 0.0036 26 0.0074 0.0048 0.0080 0.0087 0,0047 Since both turbidity and 0.0086 0.0094 0.0040 0.0027 27 0.0148 0.0080 0.0113 0.0045 0.0165 0.0038 0.0133 0.0060 0.0037 coloring matter absorb light 0.0093 0.0187 0.0060 0.0223 0.0067 28 0.0220 0.0168 0.0122 0.0046 0.0083 0.0030 0.0202 0,0034 29 0.0208 0.0173 0.0184 0.0072 0.0048 selectively and differ merelv Av. 0,0177 0.0037 0.0166 0.0048 0.0188 0.0026 0.0120 0.0094 0.0035 in de,gre-e of d i s p e r s i o n , line Of=c8emarL!atian must a Results me expressed as -log t , except €67 the Landt and Witte method, where they.arsgiyen as absolute turbidity for a concentration of 100 gram of solids in 100 ml. of solution. be drawn a r b i t r a r i l y , a n d either any one of the filtraThis is done in the method of Landt and Witte by multionmethods or the method of the writers may be chosen tiplying by the fk factor, derived from theory, and in that for this purpose. This applies to raw sugars also, and in of Zerban, Sattler, and Lorge by the use of the empirical the light of the results recently obtained by the writers formulas 1 and 2 , based on experimental evidence. If the there is no need to use unfiltered solutions for turbidity two methods give the same results, when reduced to a common determination and filtered solutions for color determination basis, this constitutes strong presumptive evidence that these in raw sugars, as previously proposed, because it is now known results are correct. An examination of the figures given in that the color determined after previous filtration is no more Table 11, columns 8 and 10, shows that there is a nearly conreliable than that derived, by the method of the writers, stant ratio, averaging 3.47, between the N values and the from the Tyndall beam intensity and transmittancy of the absolute turbidities. Table 111, column 3, shows the values unfiltered solution. The method of Landt and Witte cannot obtained by multiplying the absolute turbidity by this factor be used for this purpose on raw sugars, because the assump3.47. 'The results check closely with those found directly by tions on which the fh values of these authors are based, while the method of Zerban, Sattler, and Lorge (column 2), and the TABLE111. COMPARISON BETWEEN N, AS -log t, FOUND DIaverages agree within 1.6 per cent. RECTLY AND CALCULATED FROM ABBOLUTE TURBIDITY, BY AVERThe largest percentage discrepancies between N found AGE AND BY VARIABLE RATIO directly, and N calculated from the absolute turbidity, are N N N found insamples 14 and 2. Both samples show high total Founh by Calcd: by Calod: by Zerban and Average absorption, as may be seen from Table I, column 2. Under Variable No. Battler Ratio 3.47 Deviation Ratio Deviation this condition the fk values of Landt and Witte are evidently % % a little too high, and the theoretical assumptions on which 1 0.0104 0.0103 -1.0 0.0106 $1.0 they are based no longer hold strictly. As a matter of fact, 0.0242 2 0.0287 $10.3 0.0254 $5.0 3 0.0090 0.0092 $2.2 0.0094 a closer analysis of the figures in columns 8 and 10 of Table I1 $4.4 0.0035 4 0.0033 -5.7 0.0035 0.0 shows that the ratio between N found and N calculated de6 0.0030 0.0032 $8.7 0.0033 $10.0 8 0.0190 0.0202 +8.3 0.0198 $4.2 creases from 3.70 to 3.22 with an increase in the absolute 0.0190 0.0187 -1.6 7 0.0184 -3.2 0.0059 8 0.0054 -8.5 0.0056 -5.1 turbidity from 0 to 0.01. N has been calculated from the 0.0074 9 0.0076 $2.7 0.0078 $6.4 absolute turbidity by this varying ratio also, and the resdlts 0.0134 10 0.0138 i-1.5 0.0140 $4.6 0.0050 11 0.0047 -6.0 are shown in column 5 of Table 111. The individual values as 0.0049 -2.0 0.0084 12 0.0081 -3.8 0.0084 0.0 well as the average value of N calculated in this manner check 0.0073 13 0.0073 0.0076 0.0 $2.7 0.0094 14 0.0106 $12.8 0.0108 $14.9 with the found values only slightly better than those obtained 0.0218 15 0.0226 0,0218 $4.2 +0.9 0.0137 18 0.0132 -3.8 0.0133 -2.9 by means of the average ratio 3.47, and nothing is gained in 0.0142 17 0.0140 -1.4 0.0140 -1.4 practice by using the sliding ratio for calculating N from the 0.0085 18 0.0079 -7.1 0.0081 -4.7 0.0137 19 0.0141 $2.9 0.0142 $3.8 absolute turbidity of white sugars. 0.0070 20 -1.4 0.0089 0.0071 +1.4 While the results of the method of Zerban, Sattler, and 0.0100 21 0.0098 -2.0 0 * 0100 0.0 0.0093 22 0.0098 $3.2 0.009s $5.4 Lorge are confirmed by those of the Landt and Witte method, 0.0084 23 0.0061 -4.7 0.0083 -1.6 0.0318 24 0.0341 $7.9 0.0318 $0.8 both give an exact measure of the turbidity concentration 0.0123 25 4-0.8 0.0124 0.0126 $2.4 only if the distribution of particles of different size, shape, 0.0094 28 -2.i 0.0092 0.0094 0.0 0.0133 27 0.0129 0.0130 -3.0 -2.3 etc., is always the same. I n the method of Zerban, Sattler, 0.0158 28 0.0169 $0.8 0.0159 $0.8 0.0184 29 0.0181 -1.8 0.0180 -2.4 and Lorge, the distribution must be the same as in the standard sugar used by the writers to establish their formulas Av. 0.01200 0.01219 a4.0 0.01218 a3.2

a

INDUSTRIAL AND ENGINEERING CHEMISTRY

232

holding for sugars with little or no coloring matter, do not apply when large quantities of coloring matter are present. The experiments of the writers have shown that even with white sugars f k is slightly too high when absorption is high, and that with dark raw sugars it becomes so much too high that absolute turbidities greatly in excess of 1 are found, which is an impossibility, since the intensity of the scattered or reflected light cannot be greater than that of the incident light. But the objections to the method of Landt and Witte in the case of raw sugars do not apply to that of Zerban, Sattler, and Lorge, which is based on experiments with high concentrations of both turbidity and coloring matter, and is therefore applicable to raw as well as to white sugars. I n this simplified method, without any filtration whatever, the "quality index" (6) loses its meaning, being equal to 1 in all cases. The nature of the turbidity may be characterized by making measurements a t different wave lengths and calculating the dispersion quotients; that of the coloring matter may be similarly defined by the Q- or R-ratios of Peters. Further refinements may be introduced by independent measurement of the various particleproperties, but for routine purposes this would hardly be necessary. The suggestion is again advanced that the International Commission for Uniform Methods of Sugar Analysis undertake a comparative study of the various methods proposed and recommend one for general adoption.

Summary and Conclusions The previous investigation on raw cane sugars has been extended to hard refined sugars. A comparison has been made between the method of Balch, in which the turbidity and

VOL. 9, NO. 5

coloring matter are found from the transmittancy of the solution before and after filtration, and the method of the writers, in which both are calculated from the transmittancy and TyndaU beam intensity of the unfiltered solution. The method of Landt and Witte for determining turbidity has also been included in this study. The previous finding, that filtering agents act selectively, has been confirmed. Generally, Celite gives lower color values than asbestos, and silica gel still lower ones; the opposite is true for the turbidity figures. But there are exceptions to this rule, depending on the nature of the turbidity and of the coloring matter. The method of the writers usually gives higher color values than any of the filtration methods. While this also indicates that the filtering agents remove coloring matter, i t must be considered that both turbidity and coloring matter absorb light selectively and that any line of demarcation between the two must necessarily be arbitrary. The choice between the two types of methods thus becomes a matter of agreement. The method without filtration has the advantage of greater speed, but requires greater care in manipulation. It is suggested that the decision be left to the International Commission for Uniform Methods of Sugar Analysis.

Literature Cited (1) Balch, IND. ENQ.CHEM., Anal. Ed., 3, 124 (1931). (2) Landt and Witte, 2. Ver. deut. Zuckerind., 84, 450 (1934). (3) Zerban and Sattler, IND. ENQ.CHEM., Anal. Ed., 8 , 168 (1936). (4) Zerban, Sattler, and Lorge, Ibid., 6 , 178 (1934). (5) Ibid., 7, 167 (1935).

RECEIVED December 30, 1936. Presented before the Division of Sugar Chemistry at the 93rd Meeting of the American Chemical Society, Chapel Hill, N. C., April 12 to 15, 1937.

Estimation of Potassium in Silicates and Soils A Low-Melting Alkali Carbonate Flux J. E. GIESEKING AND H. J. SNIDER, University of Illinois, Urbana, 111.

D

IFFICULTIES in completely decomposing highly aluminous minerals by the J. Lawrence Smith fusion (2) have been pointed out by Lamar, Hazel, and O'Leary (1), and have been experienced by the present authors in working with soils, clays, and similar materials. This led to the search for ti more desirable flux, which, to be suitable for potassium determinations in silicates, (1) should have a low melting point, (2) should be reactive with silicates a t the melting point, and (3) must not leave a residue of salts difficult to separate from potassium. A mixture of two parts of sodium carbonate and one part of lithium carbonate was found to have these characteristics, if potassium is determined by the cobaltinitrite method. This flux appears to be very similar to the composition of a eutectic compound of sodium and lithium carbonates. It melts between 470" and 480" C. and is very effective in decomposing soil silicates a t temperatures between 500' and 600" C. Sodium and lithium salts do not interfere with the determination of potassium by the cobaltinitrite method. The following experimental procedure was applied to a number of soils in connection with the potassium determination: A 1-gram sample of 100-mesh soil is weighed into a 40-ml. platinum crucible. In order to deptroy the organic matter in the

sample, 0.5 to 1 ml. (depending on organic matter content) of concentrated perchloric acid is added The crucibte is placed on a hot plate and heated until the last trace of perchloric acid i s removed. To the sample, free of organic matter, are added 8 yams of a mixture of 2 parts sodium carbonate and 1 part ithium carbonate. The sample and carbonates are mixed thoroughly by means of a glass rod and then covered with an additional gram of the carbonate mixture. The crucible is covered and the sample is fused in a muffle furnace at a temperature betweeen 500' and 600" C., this temperature being maintained until all bubbling has ceased. The crucible is removed from the furnace and given a rotary motion to distribute the fused maas on the sides of the crucible d u h g solidification, thus facilitating the removal of the cake which, after cooling, is placed in a 250-ml. beaker. The cake and residue left in the crucible are dissolved together (in a covered beaker) b adding 25 ml. of water and 50 ml. of 6 N hydrochloric acid. A sxell of silica will form on the cake which must be broken off from time to time to hasten solution of the inner portion. If it dissolves too slowly the beaker may he placed on a steam bath. After the carbonates have completely dissoIved from the cake the crucible and lid are washed and removed from the beaker and the contents are taken to dryness on the steam bath. The residue is treated with 10 ml. of concentrated perchloric acid and taken to dryness on the hot plate. The perchloric acid treatment is repeated with the addition of 1.O ml. of concentratedhydrochloric acid and 0.5 ml. of concentrated nitric acid to ensure complete dehydration of the silica and decomposition of a trace of ammonium salts which might be present. The residue is treated